RSA Digital Signature

Standards
Burt Kaliski, RSA Laboratories
23rd National Information Systems
Security Conference, October 16–19, 2000
Outline

I. Background
II. Forgery and provable security
III. Contemporary signature schemes
IV. Standards strategy
Part I: Background
General Model

• A signature scheme consists of three (or more)

related operations
• Key pair generation produces a public/private key
pair
• Signature operation produces a signature for a
message with a private key
• Verification operation checks a signature with a
public key
Types of Signature Scheme

Mr = recoverable part • Appendix: message

Mnr = non-recoverable part Mr Mnr transmitted with
signature
• Total message
private key Sign recovery: message
recoverable from
signature
• Partial message
public key Verify recovery: part of
message recoverable
from signature, part
transmitted
Mr
or “invalid”
Trapdoor One-Way Functions

• A one-way function f(x) is easy to compute but

hard to invert:
– easy: x → f(x)
– hard: f(x) → x

• A trapdoor one-way function has trapdoor

information f-1 that makes it easy to invert:
– easy: f(x), f-1 → x = f-1(f(x))

• Many but not all signature schemes are based on

trapdoor OWFs
RSA Trapdoor OWF

• The RSA function is

f(x) = xe mod n
where n = pq, p and q are large random primes,
and e is relatively prime to p-1 and q-1
• This function is conjectured to be a trapdoor OWF
• Trapdoor is
f-1(x) = xd mod n
where d = e-1 mod lcm(p-1,q-1)
Embedding Operations

• An embedding operation µ(M) maps from

message strings to “message representatives,”
which can be input to f-1
– e.g., a hash function with padding
– may be randomized

• Inverse operation checks whether a message

representative is correct
– in scheme with message recovery, also recovers
message part

• Current RSA signature schemes differ primarily in

terms of the embedding operation
Scheme with Appendix

• Signature generation
M “valid” / “invalid”
embeds message,
applies trapdoor:
µ(M))
– s = f-1(µ
µ µ-1 • Signature verification
applies OWF, checks
against message:
s – µ-1(f(s), M) valid?
f-1 f
Scheme with Message Recovery

• Signature generation
Mr Mnr Mr or “invalid” embeds message,
applies trapdoor:
µ(Mr, Mnr))
– s = f-1(µ
µ µ-1 • Signature verification
applies OWF, checks
against Mnr, recovers
s M r:
f-1 f – Mr = µ-1(f(s), Mnr)
Embedding Properties

• Embedding operation should have similar

properties to a hash function:
– one-way: for random x, hard to find M s.t. µ(M) = x
– collision-resistant: hard to find M1, M2 s.t. µ(M1) = µ(M2)

• May also identify underlying algorithms

– but if so, must be done with care

• Should also interact well with trapdoor function

– ideally, mapping should appear “random”
Multiplicative Properties of RSA

• RSA function is a multiplicative homomorphism:

for all x, y,
f (xy mod n) = f(x) f(y) mod n
f-1(xy mod n) = f-1(x) f-1(y) mod n
• More generally:

f-1(∏ xi mod n) = ∏ (f-1(xi)) mod n

• Property is exploited in most forgery attacks on
RSA signatures, but also enhances recent
security proofs
Part II: Forgery and
Provable Security
Signature Forgery

• A forgery is a signature computed without the

signer’s private key
• Forgery attacks may involve interaction with the
signer: a chosen-message attack
• Forgery may produce a signature for a specified
message, or the message may be output with its
signature (existential forgery)
Multiplicative Forgery

• Based on the multiplicative properties of the RSA

function, if

µ(M) = ∏ µ(Mi)^α
αi mod n
then

σ(M) = ∏ σ(Mi)^α
αi mod n
• Signature for M can thus be forged given the
signatures for M1, …, Ml under a chosen-message
attack
Small Primes Method

• Suppose µ(M) and µ(M1), …, µ(Ml) can be factored

into small primes
– Desmedt-Odlyzko (1986); Rivest (1991 in PKCS #1)

• Then the exponents αi can be determined by

relationships among the prime factorizations
• Requires many messages if µ maps to large
integers, but effective if µ maps to small integers
• Limited applicability to current schemes
Recent Generalization

• Consider µ(M), µ(M1), …, µ(Ml) mod n, and also

allow a fixed factor
– Coron-Naccache-Stern (1999)

• Effective if µ maps to small integers mod n times

a fixed factor
• Broader applicability to current schemes:
– ISO 9796-2 [CNS99]
– ISO 9796-1 [Coppersmith-Halevi-Jutla (1999)]
– recovery of private key for Rabin-Williams variants
[Joye-Quisquater (1999)]
Integer Relations Method

• What if the equation

µ(M) = f(t) ∏ µ(Mi)^α

αi
could be solved without factoring?
• Effective for weak µ
• ISO 9796-1 broken with three chosen messages
[Grieu (1999)]
Reduction Proofs

• A reduction proof shows that inverting the

function f “reduces” to signature forgery: given a
forgery algorithm F, one can construct an
inversion algorithm I
• Provable security: inversion hard → forgery hard
• “Tight” proof closely relates hardness of
problems
Random Oracle Model

• In the random oracle model, certain functions are

considered “black boxes”: forgery algorithm
cannot look inside
– e.g., hash functions

• Model enables reduction proofs for generic

forgery algorithms — inversion algorithm hides
value to be inverted in oracle outputs
• Multiplicative properties of RSA can enhance the
proof
Part III: Contemporary
Signature Schemes
Overview

• Several popular approaches to RSA signatures

• Approaches differ primarily in the mapping µ
• Some differences also in key generation
• Some also support Rabin-Williams (even
exponent) signatures

• There are many other signature schemes based

on factoring (e.g., Fiat-Shamir, GQ, Micali, GQ2);
focus here is on those involving the RSA function
Schemes with Appendix

• Basic scheme
• ANSI X9.31
• PKCS #1 v1.5
• Bellare-Rogaway FDH
• Bellare-Rogaway PSS
• IEEE P1363a version of PSS
Basic Scheme

• µ(M) = Hash(M)
• Pedagogical design
• Insecure against multiplicative forgery for typical
hash sizes
• (Hopefully) not widely deployed
ANSI X9.31
(Digital Signatures Using Reversible Public-Key
Cryptography for the Financial Services Industry, 1998)

• µ(M) = 6b bb … bb ba || Hash(M) || 3x cc
where x = 3 for SHA-1, 1 for RIPEMD-160
• Ad hoc design
– cc octet for RW support

• Resistant to multiplicative forgery

– some moduli are more at risk, but still out of range

• Widely standardized
– IEEE 1363, ISO/IEC 14888-3
– US NIST FIPS 186-1

• ANSI X9.31 requires “strong primes”

PKCS #1 v1.5
(RSA Encryption Standard, 1991)

• µ(M) = 00 01 ff … ff 00 || HashAlgID || Hash(M)

• Ad hoc design
• Resistant to multiplicative forgery
– moduli near 2k are more at risk, but still out of range

• Widely deployed
– SSL certificates
– S/MIME

• Included in IEEE P1363a; PKCS #1 v2.0 continues

to support it
ANSI X9.31 vs. PKCS #1 v1.5

• Both are deterministic

• Both include a hash function identifier
• Both are ad hoc designs
– both resist [CNS99]/[CHJ99] attacks

• Both support RSA and RW primitives

– see IEEE P1363a contribution on PKCS #1 signatures
for discussion

• No patents have been reported to IEEE P1363 or

ANSI X9.31 for these mappings
Bellare-Rogaway FDH
(Full Domain Hashing, ACM CCCS ’93)

• µ(M) = Full-Length-Hash(m)
• Provably secure design
– resists any attack where hash function is considered a
black box, provided that RSA is hard to invert

• Variant included in IEEE P1363a, PKCS #1 v2.1

draft
Bellare-Rogaway PSS
(Probabilistic Signature Scheme, Eurocrypt ’96)

• µ(M) ≈ H || G(H) ⊕ salt

where H = Hash(salt, M), salt is random, and G is
a mask generation function
• Provably secure design
• Variant included in IEEE P1363a, PKCS #1 v2.1
draft
FDH vs. PSS

• FDH is deterministic, PSS is probabilistic

• Both are provably secure designs
– same paradigm as Optimal Asymmetric Encryption
Padding (OAEP)

• PSS has tighter security proof, is less dependent

on security of hash function
• PSS-R variant supports message recovery, partial
message recovery
• PSS is patent pending (but generously licensed)
IEEE P1363a Version of PSS

• µ(M) = G(H) ⊕ [00 … 01 || salt] || H || bc

where H ≈ Hash(salt, Hash(M)), salt is random,
and G is a mask generation function
• Salt combined with Hash(M) rather than M for
practical and security reasons:
– “single-pass” processing
– provable security if Hash(M) outside crypto module
– protection against fault-analysis attacks

• Salt can be omitted for FDH-like scheme

Schemes with Message Recovery

• Basic scheme
• ISO/IEC 9796-1
• ISO/IEC 9796-2
• Bellare-Rogaway PSS-R
• IEEE P1363a version of PSS-R
Basic Scheme

• µ(Mr) = Mr
• Another pedagogical design (“textbook RSA”)
• Insecure against various forgeries, including
existential forgery
– attacker can select signature s then “recover” Mr = f(s)

• Again, hopefully not widely deployed

ISO/IEC 9796-1
(Digital Signature Scheme Giving Message Recovery, 1991)

• µ(Mr) = π*(ml-1) π’(ml-2) ml-1 ml-2

π (ml-3) π(ml-4) ml-3 ml-4 ...
π(m3) π(m2) m3 m2
π(m1) π(m0) m0 6
where mi is the ith nibble of Mr and π*, π’ and π are
permutations
• Ad hoc design with significant rationale
• Not resistant to multiplicative forgery [CHJ99]
[Grieu 1999]
– may still be appropriate if applied to a hash value

• Moderately standardized
ISO/IEC 9796-2
(Digital Signature Scheme Giving Message Recovery —
Mechanisms Using a Hash Function, 1997)

• µ(Mr, Mnr) ≈ 6a || Mr || H || bc
• µ(Mr) = 4b bb … bb ba || Mr || H || bc
where H = Hash (Mr, Mnr) or Hash (Mr)
– (assumes modulus length is multiple of 8)
– general format allows hash algorithm ID

• Ad hoc design
• Not resistant to multiplicative forgery if hash
value is 64 bits or less [CNS99]
– may still be appropriate for larger hash values

• Newly standardized
Bellare-Rogaway PSS-R
(Probabilistic Signature Scheme with Recovery, 1996)

• µ(Mr, Mnr) ≈ H || G(H) ⊕ [salt || Mr]

where H = Hash(salt, Mr, Mnr), salt is random, and
G is a mask generation function
• Provably secure design
• Variant included in IEEE P1363a, draft revision of
ISO/IEC 9796-2
IEEE P1363a Version of PSS-R

• µ(Mr, Mnr) = G(H) ⊕ [00 … 01 || Mr || salt] || H || bc

where H ≈ Hash(salt, Mr, Hash(Mnr)), salt is
random, and G is a mask generation function
• Extension of PSS variant
– PSS variant is special case where Mr is null
Part IV: Standards
Strategy
Standards vs. Theory vs. Practice

• ANSI X9.31 is widely standardized

• PSS is widely considered secure
• PKCS #1 v1.5 is widely deployed

• How to harmonize signature schemes?

– (primary question for signature schemes with
appendix; related question for message recovery)
Challenges

• Infrastructure changes take time

– particularly on the user side

• ANSI X9.31 is more than just another encoding

method, also specifies “strong primes”
– a controversial topic

• Many communities involved

– formal standards bodies, IETF, browser vendors,
certificate authorities
Prudent Security

• What if a weakness were found in ANSI X9.31 or

PKCS #1 v1.5 signatures?
– no proof of security, though designs are well motivated,
supported by analysis
– would be surprising — but so were vulnerabilities in
ISO/IEC 9796-1,-2

• PSS embodies “best practices,” prudent to

improve over time
Proposed Strategy

• Short term (1-2 years): Support both PKCS #1

v1.5 and ANSI X9.31 signatures for
interoperability
– e.g., in IETF profiles, FIPS validation
– FIPS 186-2 schedule allows PKCS #1 v1.5 for an 18-
month transition period, FPKI TWG is requesting a
further extension

• Long term (2-5 years): Move toward PSS

– upgrade in due course — e.g., with AES algorithm, new
hash functions
– separate assurance requirements from interoperability
• e.g., key sizes, key protection, “strong primes”
Standards Work

• PSS, PSS-R standardization work in progress in

various forums:
– IEEE P1363a
– PKCS #1 v2.1
– ISO/IEC 9796-2 revision

• Coordination ongoing, ballot target Spring 2001

• Promotion in other forums planned
– ANSI X9.31
– FIPS
– IETF
Conclusions

• Several signature schemes based on RSA

algorithm
– varying attributes: standards, theory, practice

• Recent forgery results on certain schemes,

security proofs on others
• PSS a prudent choice for long-term security,
harmonization of standards